REPL arithmetic gives wrong result on first run


#1

REPL gives 2 different results for simple arithmetic calculation

Gives came after running a bunch of complex, quantitative code
I have not replicated it, nor tried due to lack of time

So this is just to put out there that something is not 100%

Not even sure whether its julia or Atom or Juno or what


#2

There’s really not much we can do with this. Without any clue as to what you were doing before or how to reproduce, this could be a bug in just about anything, or even something like some subtle type piracy in a library you’re using.

Can you give any detail? Or find any reproduction? Otherwise I might as well just assume your RAM was hit by a stray cosmic ray.

Also, please post code as text, not as a screenshot.


#3

Yes, I realise its pretty useless without a reproduction
I thought I would still submit in case anyone else ever runs into something like this and perhaps gives another clue
If I get it again, I will add info


#4

I will certainly keep an eye out myself :slightly_smiling_face:


#5

I’m seem to recall some Juno/Atom bug along these lines being reported previously.


#6

I’ve seen some similar strangeness (expressions that should be deterministic returning different results when run back-to-back), but also have never been able to come up with a repro. It’s only been when running from with Atom, but I also usually run within Atom, so I can’t say it wouldn’t happen at a bare REPL.

I thought I filed an issue with atom-julia-client at some point, but maybe not. Here are a couple issues I filed with FixedPointNumbers, which was the package I was hacking on when this was happening to me before:


There was also a discussion on Slack a while ago where it happened to someone else.


#7

If you update to a somewhat recent version of Atom.jl (current is 0.6.14) that should not happen any more.